Design research & practice
Praneet Koppula · 18+ years across technical product design, research, design leadership, and AI-assisted workflows.
Interactive demonstrations of concepts from Praneet’s practice. Each uses fixed sample data and is clearly labeled as a portfolio demonstration.
Choose an intent format, then step through the workflow. Each stage reveals what happens and who is involved. Human confirmation is required before execution.
This stage demonstrates parsing the selected format and building a structured model. In a production system, the model would be presented for review before any action is taken.
Shows intent captureThis is a portfolio demonstration. It does not connect to any live model or production system.
Explore the state dimensions of a simulation slicing session. Each tab reveals how inspectable state helps engineers understand and trust complex tools. Based on concepts from Model Slicer work.
The data or signal the system operates on. Knowing what the tool sees is the first step in trusting what it produces.
Why this matters: When you know exactly what went in, you can assess whether the output is reasonable for the given input.
What the analysis covers and what it deliberately excludes. Explicit boundaries prevent silent overreach.
Why this matters: A tool that shows its scope helps engineers avoid relying on results that may not apply to their specific scenario.
Where the data originated and how it was collected. Provenance at the source level enables reproducibility.
Why this matters: Knowing the data source lets engineers verify conditions and reproduce results independently.
The temporal slice under analysis. Time windows allow engineers to focus on specific event sequences without reprocessing the entire dataset.
Why this matters: Slicing time lets engineers isolate specific behaviors without losing the surrounding context.
What the system produced within the defined scope and constraints. Results are always presented alongside their context, never in isolation.
Why this matters: Results without context are misleading. A highlighted slice tells the engineer precisely what the tool found relevant.
The record of how this slice was created: from what parent analysis, and what parameters were used. Every analysis has a lineage.
Why this matters: Provenance creates accountability. Engineers can trace results back to specific configurations and replicate or challenge them.
Whether the current slice is still meaningful given changes in the underlying model or constraints. Validity has a shelf life.
Why this matters: A valid result from a stale configuration is worse than no result. Validity tracking tells engineers when to regenerate.
What happens when constraints change, data drifts, or the slice becomes invalid. Recovery paths are a first-class part of the state model, not an afterthought.
Why this matters: Systems that handle recovery gracefully earn more trust than systems that silently fail or require manual reset.
This is a portfolio demonstration using representative sample data. Not connected to any production system or internal study.
Select a principle, review the interface mockup, assign a rating, and compare with a design reviewer’s perspective. Inspired by Praneet’s design-principles review practice.
Training Configuration
This is a portfolio demonstration. Ratings are illustrative and do not represent user research or production analytics.
Three projects that shaped my approach to designing complex, trustworthy systems.
I led UX vision with an engineering lead for an AI assistant spanning requirements, architecture, implementation, simulation, navigation, and explanation. The interaction emphasized traceability, preview, confirmation, and human control at every stage.
Case study ↗I improved early productization and later workflow concepts for dynamic slicing, simulation data, constraints, time windows, provenance, recalculation, validity, and failure states. The work demonstrated why inspectable state helps users trust complex engineering tools.
Public context ↗I developed a shared workflow language and reusable design-principles review practice. The process uses independent ratings followed by facilitated discussion, coaching, and lightweight enablement to help teams scale judgment without losing individual perspective.
Case study ↗I managed distributed UX leadership across 4 countries, working within a broader governance ecosystem spanning 120 products. My work combined strategic leadership development with hands-on contribution to workflow and interaction direction.
I developed UX design and research leaders while remaining directly involved in the craft: shaping how complex tools reveal their state, how AI-assisted workflows stay accountable to human judgment, and how design teams build shared language around principles.